OBJECTIVE: This study examines one-, two-, and three-factor models of attention-deficit/hyperactivity disorder (ADHD) using the existing 18 Diagnostic and Statistical Manual of Mental Disorder, 4th edition (DSM-IV) symptoms in a sample of symptomatic preschoolers. METHODS: Parent and/or teacher ratings of DSM-IV symptoms were obtained for 532 children (aged 3-5.5) who were screened for the Preschool ADHD Treatment Study (PATS). Confirmatory factor analysis (CFA) using symptoms identified on the Conners' Parent and Teacher Rating Scales was conducted to assess a two-factor model representing the DSM-IV dimensions of inattention (IN) and hyperactivity/impulsivity (H/I), a three-factor model reflecting inattention, hyperactivity, and impulsivity, and a single-factor model of all ADHD symptoms. Exploratory factor analysis (EFA) was subsequently used to examine the latent structure of the data. RESULTS: For parent ratings, the two-factor and three-factor models were marginally acceptable according to several widely used fit indices, whereas the one-factor model failed to meet minimum thresholds for goodness-of-fit. For teachers, none of the models was a solid fit for the data. Maximum likelihood EFAs resulted in satisfactory two and three-factor models for both parents and teachers, although all models contained several moderate cross loadings. Factor loadings were generally concordant with those published for older children and community-based samples. CONCLUSION: ADHD subtypes according to current DSM-IV specifications may not be the best descriptors of the disorder in the preschool age group.